1 / 20

Parallel Solution to the Radiative Transport

Parallel Solution to the Radiative Transport. EG PGV 2009. Szirmay-Kalos László Liktor Gábor Tam ás Umenhoffer Tóth Balázs Glenn Lupton Kumar Shree. TU Budapest. Overview. Radiative transport Challenges of parallel iteration Our approach Initial estimation Modified iteration

trevet
Download Presentation

Parallel Solution to the Radiative Transport

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Parallel Solution to the Radiative Transport EG PGV 2009 Szirmay-Kalos László Liktor Gábor Tamás Umenhoffer Tóth Balázs Glenn Lupton Kumar Shree TU Budapest

  2. Overview • Radiative transport • Challenges of parallel iteration • Our approach • Initial estimation • Modified iteration • FCC GRID • CUDA • Results

  3. Radiative transport screen Camera Outgoing radiancia: L(s+ds) Incident radiancia: L(s) path: ds Emission Absroption Out-scattering In-scattering

  4. Solution methods • Monte-Carlo simulation • O(m -0.5) • Parallelization is trivial • Iteration • O(m) • Parallelization is a challenge

  5. Iteration • Finite-element approaches (grid) • Iteratively refines the estimation • Error depends on the initial guess L = TL + Q Ln= TLn-1 + Q ||Ln-L||<n ||L0-L||

  6. Parallel Iteration Boundary affecting multiple blocks • Costly data exchanges • Less frequent data exchanges • Few iteration steps: Good initial guess • Unscattered component • Homogeneous solution • Approximate inhomogeneous node1 node2 node3 node4

  7. Initial approximation • Direct term • Direct + Indirect • Solve diff. equation for each ray assuming spherical symmetry

  8. Iteration refinement - Finite element: FCC grid FCC BCC CC

  9. Reduced data exchanges node1 TLn-1+Q Ln-1 Ln node2 Ln1 Ln-11 T1 node3 T12 Ln2 T21 T2 Ln-12 Ln-22 node4

  10. Reduced data exchanges + TLn-1+Q TLn-2+Q TLn-2+Q Ln T[T12](Ln-3-Ln-2) Noise converges to zero!

  11. Iteration solution: CUDA • Sampling

  12. Iteration solution: CUDA • Sampling • Illumination network

  13. Iteration solution: CUDA • Sampling • Illumination network • Initial radiance distribution

  14. Iteration solution: CUDA • Sampling • Illumination network • Initial radiance distribution • Iteration

  15. Iteration solution: CUDA • Sampling • Illumination network • Initial radiance distribution • Iteration • Visualization

  16. Visualization: 5 node HP SVA node 1 node 2 …

  17. Error analysis for the initial distribution

  18. Scalability Error 2% Single iteration Compute + Communication

  19. Results Direct term 25 iterations 100 iterations Direct+Indirect estimation

  20. Conclusions • Interactive solution of the radiation transport • Scalable iteration scheme • Current limitations • No specular reflections • Point sources

More Related